Mercury levels in hair are associated with reduced neurobehavioral performance and altered brain structures in young adults

Mercury is a neurotoxic heavy metal with well-documented severe effects at high exposure and during in utero development. In the general population, methylmercury exposure mainly occurs via fish consumption. Prior epidemiological findings on low-level adult exposure are mixed: some studies report poorer cognition (especially processing speed), others find no association, and some report lower odds of depression with higher blood mercury. Mechanistically, methylmercury can affect myelin/axons, induce oxidative stress, disrupt cytoskeletal elements, and alter neurotransmission. Three gaps motivated this study: (1) lack of research on low-level mercury associations with brain morphometry using advanced MRI in adults, (2) inconsistent adult findings possibly due to a mature blood–brain barrier reducing CNS accumulation, and (3) reliance on blood/urine mercury (sensitive to short-term fluctuation) rather than hair mercury, which reflects longer-term exposure. The study aimed to test whether hair mercury levels in young adults relate to cognitive performance, depressive tendency, and brain structural metrics (rGMV, rWMV, FA, MD), hypothesizing higher mercury would associate with reduced cognition (especially processing speed), higher depression, lower rGMV/rWMV, lower FA, and higher MD.

The authors review evidence that high-level mercury exposure impairs neurocognition and development. Several studies link higher blood mercury to lower cognitive performance in adults and adverse effects of maternal mercury on offspring cognition, though some adult studies show null effects. Depression findings are inconsistent, with a large cross-sectional study noting higher blood mercury associated with lower odds of depression in older adults, while others found no link with amalgam/urinary mercury. Animal and basic studies highlight mercury’s neurotoxicity, including effects on myelin/axons and diverse mechanisms such as oxidative stress, cytoskeletal disruption, and neurotransmitter system alterations. Methodologically, hair mercury is argued to be a preferred biomonitoring matrix for longer-term exposure, less affected by short-term intake variations than blood/urine.

Design and participants: Cross-sectional analysis of 920 healthy, right-handed young adults in Japan (561 males, 359 females; mean age 20.7 years, range 18–27), mostly university students/graduates. Exclusion included neurological/psychiatric illness, MRI contraindications, and drug use. Ethics approval and informed consent obtained. Hair mercury assessment: Scalp hair samples (~4 cm length, 0.1 g) were collected near the scalp. Samples (~75 mg) were washed (acetone, 0.01% Triton), digested in 6.25% tetramethylammonium hydroxide with gold stabilizer, and analyzed by ICP-MS (Agilent-7500ce) with internal standards (Sc, Ga, In). Certified reference hair material (NIES CRM no. 13) was used for QC. Results expressed as µg/g (ppm). Hair mercury was log-transformed for analyses to normalize distribution. Hair treatment history (dye/perm/bleach) was recorded; perming showed only weak correlation with Hg and was not included as a covariate in main models. Correlations with other hair trace elements were examined; adding arsenic did not materially change results. Neuropsychological measures: A comprehensive battery emphasized processing speed and mood: Raven’s Advanced Progressive Matrices (general fluid intelligence); Tanaka B-type Intelligence Test (TBIT) Type 3B total and three factors (Perception/simple processing speed; Spatial relations; Reasoning); timed arithmetic (simple and complex multiplication); Stroop (Hakoda’s version) including Word-Color and Color-Word tasks (processing speed), plus Stroop and reverse-Stroop (inhibition); Japanese reading comprehension; S-A creativity test (divergent thinking); computerized digit span (forward+backward sum). Depression was assessed with the Japanese Beck Depression Inventory (BDI). MRI acquisition: 3T Philips Achieva. T1-weighted MPRAGE: 240×240 matrix, TR=6.5 ms, TE=3 ms, FOV 24 cm, 162 slices, 1.0 mm thickness. Diffusion-weighted imaging: spin-echo EPI, TR=10293 ms, TE=55 ms, FOV 22.4 cm, 2×2×2 mm3 voxels, 60 slices, 32 directions (b=1000 s/mm2) plus three b0 images. FA and MD maps computed on the scanner console (with motion/eddy-current correction). Structural preprocessing (VBM): SPM12 new segmentation (6 tissues; Thorough Clean; East Asian template; sampling distance 1 mm), DARTEL for high-dimensional registration (template from 800 participants), modulation with Jacobian determinants, normalization to MNI space (1.5 mm isotropic), and smoothing with 8-mm FWHM. rGMV and rWMV derived for voxelwise analyses. DTI preprocessing: SPM8 two-step segmentation using FA and MD to create synthesized images enabling robust GM/WM segmentation; DARTEL-based normalization (1.5 mm isotropic). FA analysis constrained to a WM mask (WM probability >0.99) and smoothed 6-mm FWHM; MD analysis constrained to GM/WM mask (GM+WM probability >0.99) and smoothed 8-mm FWHM. Quality control by visual inspection; artifacts excluded. Supplemental TBSS analyses were also performed for FA (results consistent in trend but mostly not significant at voxelwise level; ROI skeleton means correlated with hair Hg). Covariates and statistical analysis: Psychological outcomes analyzed by partial correlations controlling for sex, age, self-reported height, weight, BMI, annual family income, parents’ highest educational qualifications, fatty fish intake, and intake of less-fatty fish. Multiple comparisons across 15 psychological tests were controlled using FDR (graphically sharpened method), two-sided alpha 0.05. Whole-brain multiple regression analyses (SPM8) tested associations of hair Hg with rGMV, rWMV, FA, and MD, covarying the same factors (age, sex, height, weight, BMI, family income, parents’ education, fatty fish intake, less-fatty fish intake). Analyses were masked appropriately (GM/WM thresholds noted). Multiple comparison control used TFCE with 5000 permutations and FWE-corrected P<0.05 at the voxel level (one-sided in SPM). Sample sizes: VBM N=920; DTI N=919. Sensitivity analyses excluding fish intake covariates were conducted; results were broadly similar though some volume effects attenuated, likely due to positive correlation of fish intake with brain volume and hair Hg.

Sample characteristics: 920 participants; mean age 20.7 years. Average hair Hg levels comparable to previous modern Japanese cohorts and higher than in lower-fish-intake countries. Psychological associations (partial correlations, adjusted; FDR-corrected): Higher log hair Hg was associated with lower cognitive performance on speeded tasks and lower depressive symptoms. Significant results included: - TBIT Total: r = -0.080, p(unc)=0.018, p(FDR)=0.034. - TBIT Perception factor (simple processing speed): r = -0.075, p(unc)=0.028, p(FDR)=0.034. - Word-Color task: r = -0.079, p(unc)=0.017, p(FDR)=0.034. - Color-Word task: r = -0.074, p(unc)=0.026, p(FDR)=0.034. - Beck Depression Inventory: r = -0.087, p(unc)=0.008, p(FDR)=0.034 (i.e., higher Hg associated with lower depressive tendency). Effect sizes were small (|r|<0.09). Other measures were not significant after FDR. rGMV: Higher hair Hg associated with lower rGMV in a cluster mainly in the left thalamus extending into the left hippocampus (MNI x,y,z = -9, -25.5, 4.5; T=4.63; TFCE=1699.25; FWE-corrected p=0.031). rWMV: Higher hair Hg associated with lower rWMV in widespread bilateral white matter tracts, including cerebellar peduncles, pontine crossing tract, corticospinal tract, corpus callosum (body/splenium), lemniscus, fornix and cingulum, internal/external capsule, corona radiata (anterior/superior/posterior), posterior thalamic radiation, sagittal stratum, stria terminalis, superior/inferior fronto-occipital fasciculi, superior longitudinal fasciculus, uncinate fasciculus, and tapetum. Effect sizes for mean cluster values were small (|partial r|<0.140). FA: Contrary to hypothesis, higher hair Hg was associated with greater FA in bilateral WM clusters overlapping with regions of rWMV reduction. Significant clusters (TFCE FWE-corrected): - Left posterior regions (posterior limb/retrolenticular internal capsule, posterior/superior corona radiata, posterior thalamic radiation, superior longitudinal fasciculus, tapetum): peak at (-36, -45, 7.5), T=4.48, FWE p=0.004, cluster size 889. - Right middle WM regions (internal/external capsule, superior/posterior corona radiata, posterior thalamic radiation, superior longitudinal fasciculus): peak at (33, -12, 27), T=4.14, FWE p=0.005, cluster size 1743. - Splenium/posterior corona radiata: peak at (19.5, -42, 30), T=3.85, FWE p=0.022, cluster size 250. MD: Higher hair Hg was associated with lower MD in extensive right-lateralized frontal, cingulate, basal ganglia, and temporal regions, and additional clusters in left anterior frontal and right posterior cingulum/superior parietal/angular gyrus. Notable gray/white matter inclusions: right anterior/middle cingulate, supplementary motor area, right caudate/putamen/pallidum, right frontal opercular/orbital/medial regions, insula, right temporal pole/amygdala/anterior temporal gyrus, and right superior parietal/angular regions; with involvement of WM tracts such as anterior limb of internal capsule, corona radiata, external capsule, superior/inferior fronto-occipital fasciculi, cingulum, genu/body/splenium of corpus callosum. Representative cluster statistics: largest cluster TFCE=1569.00, FWE p=0.010, cluster size 11576 voxels; additional clusters FWE p=0.032–0.050 (see Table 4 in text). Overlap with rWMV/FA effects was limited, with partial overlap in WM between basal ganglia and frontal lobe. Post-hoc brain–behavior associations: No significant associations after FDR between Hg-linked brain clusters and Hg-linked psychological measures; uncorrected trends were consistent with prior reports linking TBIT with hippocampal rGMV and processing speed with rWMV. Overall, effects were statistically significant but small in magnitude across modalities.

Findings address the central question by demonstrating that higher long-term mercury exposure (indexed by hair levels) in young adults relates to modest decrements in processing speed, alterations in mood (lower depressive symptoms), and measurable brain structural differences: reduced rGMV in thalamus/hippocampus, widespread reductions in rWMV, increased FA in overlapping WM, and decreased MD in extensive GM/WM regions. The cognitive results align with prior reports of speed-related deficits, while the mood finding contrasts with the initial hypothesis but is consistent with some large-scale epidemiological observations of lower depression odds at higher mercury levels. Mechanistic interpretations include oxidative stress vulnerability of the hippocampus (potentially explaining lower rGMV), mercury-induced cytoskeletal disassembly and impaired synthesis leading to volume reductions, and increased FA possibly reflecting loss of intra-axonal structures (microtubules/neurofilaments) that constrain longitudinal diffusion, as shown in animal methylmercury models. The unexpected lower MD with higher Hg may reflect reduced cerebral blood flow or monoaminergic changes affecting microstructure, particularly in dopaminergic projection areas (e.g., basal ganglia). Collectively, results suggest that even low-level, chronic mercury exposure typical of high fish-consuming populations is associated with subtle but detectable neural and behavioral differences.

Using hair mercury as a long-term exposure index in a large cohort of young adults, the study shows that higher mercury levels are weakly associated with: (a) lower rGMV in left thalamus/hippocampus, (b) widespread lower rWMV, (c) higher FA in bilateral WM tracts overlapping rWMV reductions, (d) lower MD in extensive GM/WM (notably frontal lobes and right basal ganglia), (e) poorer processing speed-related cognitive performance, and (f) lower depressive tendency. These converging structural and behavioral associations indicate that even low-level mercury exposure is linked to subtle brain and neurobehavioral differences. Future work should include longitudinal and mechanistic studies, consider broader and more diverse populations, incorporate direct measures of cerebral blood flow and neurotransmitter function, and disentangle mercury effects from dietary factors (e.g., fish species/nutrients) to clarify causality and pathways.

Generalizability is limited by the homogeneous sample of well-educated young adults. The cross-sectional design precludes causal inference. Residual confounding is possible despite adjustment (e.g., detailed fish consumption patterns, nutrient co-exposures such as omega-3s/selenium, hair treatments). Effect sizes were small. Measurement error in hair trace elements and psychological outcomes may attenuate associations. Imaging analyses, while corrected for multiple comparisons, can overestimate effect sizes; preprocessing choices (SPM8/12 differences) and methodological considerations (e.g., TBSS vs. SPM pipelines) were explored but remain potential sources of variability.